[study] BERNARDINI2024_TIRE (PRJNA856813)

[sr320]

Claimed by aree start-study on Hete, 2026-10-12. Run and harmonization finished 2026-10-12. Blocked on sr320/AREE#24 (--tag_seq) and #31 (toxicant_tolerance). This branch contains both, so check-study-scope fails until they merge and it is rebased onto main.

Study

• study_id: BERNARDINI2024_TIRE. AREE's first real pollutant study.
• Accession: PRJNA856813. The paper is Bernardini et al. 2024, J Hazard Mater 480:135742, 10.1016/j.jhazmat.2024.135742. It's not open access; only the abstract was read.
• Design: adult gill on day 41. Tire microparticles (MR) or their leachate (LEA), each at low and high concentration, vs unexposed controls (CTL). One oyster per library, 6 per group, 30 libraries used. NAT is left out because no source defines it.
• Mode: raw_reanalysis, --single_end --tag_seq, native run on Hete.

Library type was checked, not assumed

The deposit says only "RNA-Seq, RANDOM". Two checks agree that these are 3′ QuantSeq FWD libraries:

• Poly-A screen: 7.4% of reads contain 15+ A's, against 0.1% for full-length RNA-seq and 15.9% for known QuantSeq.
• Strandedness: Salmon detected every library as SF (forward-stranded).

So counts are not length-corrected. An earlier AREE-co reanalysis had treated them as full-length.

Result

Comparison	n	padj < 0.05
tire particles high vs CTL (meta_analysis_primary)	6 v 6	13 / 22,199
tire particles low vs CTL	6 v 6	17 / 22,300
leachate high vs CTL	6 v 6	48 / 22,767
leachate low vs CTL	6 v 6	88 / 22,486

• Replicates are real: dispersion is 0.030–0.054, typical of individual oysters.
• Mapping: 43.5–56.7%, as expected for 3′ tags.
• Agrees with the paper: responses are small and larger for leachate, matching the abstract's activation "in the LEA treatment" and its "possible oyster resilience".
• Harmonized: 89,752 evidence records.

Checklist

• aree validate-study passes (expected stress_response override warnings)
• Sample sheet from aree fetch-samplesheet. The design is parsed from run-level experiment_title, because the BioSamples are shared 16S records; the platform is part of the condition, so 16S runs never enter an arm.
• All 30 FASTQ MD5s verified against ENA
• Manifests (with single_end / tag_seq), standardized tables and reports committed
• qc_status: passed_with_warnings, analysis_status: harmonized
• Curation tests in tests/test_bernardini2024_tire.py (7; full suite 354)
• No edits to shared code. Rebase after Support single-end and 3' Tag-seq RNA-seq, and designs from authors' tables #24 and Add a toxicant_tolerance phenotype term (ontology 0.3.0) #31 merge.

Curation decisions

• Classification: stress_response on toxicant_tolerance. No tolerant-vs-sensitive contrast exists; the abstract's "resilience" is an absence of effect at group level.
• Pooling: high-concentration tire particles is the primary. The other three share its controls.
• Unknowns left null: concentrations and tank layout. The paper wasn't read in full, so oysters may be nested in one tank per treatment.

🤖 Generated with Claude Code